The Cbir Key Technology

Along with the arrival of multimedia and information time, people can get more and more information. How to provide a efficient way to get the information which we need in deed? Information retrieval is getting more and more important. Especially, because of the large scale of images, researchers are paying out their hard work to image retrieval. Nearby, technology of Content Based Image Retrieval (CBIR) appears.There are three levels about image retrieval: perceptive level, cognitive level, and affective level. It is very difficult to deal with the cognitive level and affective level. Most research about CBIR pays attention to the perceptive level and this bring out the semantic gap between vision feature and semantics. It is very difficult to resolve the semantic gap completely in the broad fields and deeply depend on the progress of other research fields. Now, we should pay attention to vision feature exaction, retrieval algorithm and relevance feedback to reduce the semantic gap to the best of our abilities.We discuss the three aspects above in this paper and describe a novel CBIR model which pay attention to both vision feature and semantics. Also, we propose a novel idea of improved clustering which can improve the result of image retrieval. In the relevance feedback, Support Vector Machine(SVM) is used to classify images and produce the semantics. Validity of our thoughts have been testified by the experiments. But it can not resolve the semantic gap completely. With the development of other research fields, CBIR will make more and more progress in the future.